We assess the storage capacity of a solar thermal plant needed to achieve a given probability of avoiding an unscheduled reduction in output throughout the year, due to the variability of the solar resource. To our knowledge, this is the first time such an assessment has been reported. The majority of cases are assessed for the reference case of a constant output at full load, although a limited assessment of the influence of turn-down is also reported. The model accounts for the diurnal, seasonal and weather-based variability of the solar resource by the use of ten-year historical time-series of direct normal irradiance. Calculations are performed using a simplified, in-house model of the solar thermal plant that assumes stead-state at each hourly time-step. The reference conditions are selected to be best case-efficiencies for plant sub-systems. Sensitivity analyses are used to assess the influence of reduced device efficiency and of the size of the power-block relative to the heliostat field. The current analysis is independent of the type of storage, so that it assesses the quantity of storage required regardless whether it is sensible, chemical, or latent. Calculations are performed for six sites, three in the USA and three in Australia, all with an excellent solar resource. The results find that typically four to ten days of storage, together with a power-block size an order of magnitude lower than that of the maximum heliostat capacity, is required to achieve a probability of avoiding an unscheduled reduction in output from the maximum of less than one day every three years. However, the storage capacity can be reduced by a factor of five if it is acceptable to turn the output down to 40% of the maximum value. Assessments of utilisation factors and dumping energy when storage capacity is reached and solar efficiency provide further insight into these findings. (C) 2015 Elsevier Ltd. All rights reserved.